Last year what many had previously deemed unthinkable occurred. A machine mind beat a human Grandmaster at the ancient Chinese intuition based strategy game GO. A game that has been called the most complicated two person game ever invented, and one that even many AI experts had long declared would still maintain human superiority for many years to come.
Now Google has just announced to the world that same artificial intelligence, AlphaGo, has become even smarter. Though the real headline here is that it has accomplished this entirely through self-taught learning. Which leaves us with an interesting new reality as a species. That one of the smartest minds on the planet, able to out-think some incredibly smart people, is both not human nor even educated by human hands. This last claim was just recently backed up in the science journal Nature. Which over the weekend published an article featuring the machines original designers stating that this new model of AlphaGo unlike its predecessor has learned with absolutely, “no human input.”
In just a few days time it taught itself the rules of a game that has more possible board configurations than there are atoms in the Universe, before going on to beat the number one ranked human player in the world Ke Jie 3 games out of 3. DeepMind the Google owned AI company that created the machine said it learned only from interacting with itself, on a reward/punishment based system. Unlike previous attempts at AI, this new updated AlphaGo “is no longer constrained by the limits of human knowledge” at least according to Deepmind CEO Demis Hassabis.
This newly updated AI dubbed AlphaGo Zero, also is notable for using a single machine brain that more closely functions like our own human one in design and operation. Mimicking a biological “neural network” instead of the previously used multiple-computer linked machine mind the previous model and most AI still use today.
While most experts say that as long as we continue to only be able to make machines who are experts in very specific tasks we should not feel threatened, as they are still far away from the general purpose intelligence we humans possess. Now though that the machines are learning at their own pace, and by their own rules, can we be so sure of just how quickly further AI barriers will come tumbling down. Maybe tech mogul Elon Musk was onto something when he warned us all of the impending dangers present in continued unrestricted AI research.
If you enjoyed this article please share and follow @Jack_Kenrick. If you want to help ensure more similar content, please consider clicking this button and supporting directly. Everything helps